View the original community article here

Last tested: Sep 6, 2019

It's not possible to choose a visual drilling type via the UI, but there are implicit rules that say what the visualization will be. There is a feature request requesting ability to change them at will.

Workaround

You can use the method in this Help Center article to hard code the viz type for drilling on a field: https://help.looker.com/hc/en-us/articles/360023589613--More-Powerful-Data-Drilling

The Rules

1. if a dimension is the primary key OR (the first column is a dimension AND it’s name is id) OR (the first dimension’s name is _id)
   1. if there is only 1 record then “single_record” viz
   2. otherwise we’ll choose a “table”
2. if the first dimension contains “albersUsa” data and there’s only 1 measure column then the type should be “looker_geo_choropleth”
3. if any of the dimension contains location data then go for “looker_map”
4. if any of the dimension contains map layer data and any measures/dimensions are numeric then go for “looker_map”
5. if there’s no dimensions or measures then go for “looker_column” type
6. if there is only 1 row and 1 column then “single_value”
7. if there is 1 dimension and the first dimension is numeric and there is 1 or more measures then go for a “looker_line”
8. if there is 1 dimension and 1 measure and that dimension is of type “zipcode” display “looker_geo_coordinates”
9. If there is 1 or more measures and only 1 dimension which contains timeframe data
   1. if the first dimension is a time interval
      1. if the whole data set is smaller then 50 rows then type is “looker_area” and stacking is “normal”
      2. otherwise it’s “looker_column” and stacking is “normal”
   2. otherwise it’s a “looker_line” no stacking
10. if there are 2 dimensions and no measures and the 2 dimensions are a numeric type and timeframe type then “looker_scatter”

OTHERWISE (none of the conditions above are met)

• If there is more then 1 dimension OR the data is larger then 150 rows OR (there’s only 1 dimensions and no measures) then the viz is “table” type
• otherwise the type is “looker_column”